What hardware do you need to run Beam 1.0 501B-A23B?
Beam 1.0 501B-A23B at UD-Q4_K_XL is 307 GB of weights, with a context ceiling of 1M tokens. Reflection's first open-weights model: 501B total with 23.4B active, built for reasoning, software engineering and long agent runs, and predominantly for English.
How good is Beam 1.0 501B-A23B, really?
On the Artificial Analysis Intelligence Index v4.3 it scores 30 (token-efficient: 79M output tokens over the index against a median of 130M), which puts it in the Sonnet-class band. In the same league as the labs' mainstream models on this index. See the whole table.
Head to head with its neighbours on the leaderboard: vs Qwen3.8 27B above it, vs Inkling Small below.
What Beam 1.0 501B-A23B costs either way
Renting the same model costs $0.25 per million input tokens and $1 per million output (OpenRouter, cheapest of five active endpoints: Nebius at FP8, checked 2026-10-22). Buying a machine only beats that if you use it hard enough, for long enough, that the hardware price divides down below the rental bill. What a million tokens costs each way puts the two prices side by side.
The machines that miss Beam 1.0 501B-A23B, and what they run
None of the machines here with a published price that are still sold holds Beam 1.0 501B-A23B at any window, and none of them misses by a window: its 307 GB of weights are larger than the usable memory in every one of them, before a single token of key-value cache. The Mac Studio M5 Ultra, 256GB comes nearest, 114.8 GB short at 4k, the shortest window on the list. Qwen3.8 27B is the strongest model six of the eight hold, and it scores 34 on the intelligence index against this model's 30.
Two machines here do hold Beam 1.0 501B-A23B at 32k, and neither of them is an answer to what to buy, which is the question the rest of this page asks: the Mac Studio M3 Ultra, 512GB, no longer sold, at the $9,499 it launched at, and the Mac Studio M5 Ultra, 512GB, on sale with no published price. They are sold with 512 GB of memory and hand a model 384 GB of it, against the 308 GB Beam 1.0 501B-A23B needs at 32k.
| Machine | Price | Usable memory | Short by | Strongest model it holds |
|---|---|---|---|---|
| Mac Studio M5 Ultra, 256GB | $10,799 | 192 GB | 114.8 GB | GLM-5.3-Flash |
| DGX Spark GB10 Grace Blackwell, 128GB | $4,699 | 119.5 GB | 187.3 GB | Qwen3.8 27B |
| Strix Halo Framework Desktop, 128GB | $3,449 | 96 GB | 210.8 GB | Qwen3.8 27B |
| MacBook Pro M5 Max (16-inch), 128GB | $6,999 | 96 GB | 210.8 GB | Qwen3.8 27B |
| NVIDIA RTX PRO 6000 Blackwell, 96GB | $18,000card only | 95 GB | 211.8 GB | Qwen3.8 27B |
| Mac mini M5 Pro, 64GB | $2,699 | 48 GB | 258.8 GB | Qwen3.8 27B |
| AMD Radeon AI PRO R9700, 32GB | $1,299card only | 31 GB | 275.8 GB | Qwen3.8 27B |
| MacBook Air M5 (13-inch), 16GB | $1,299 | 10.5 GB | 296.3 GB | Gemma 4 12B |
One machine per family, the largest configuration in each that still misses, nearest first. Short by is the weights plus the key-value cache at 4k measured against the memory that machine's GPU can address, which that machine's own page gives. The last column is the highest-scoring current model the machine holds at 32k.
Price the Mac Studio M5 Ultra, 256GB on GLM-5.3-Flash instead
The specifics
- Parameters
- 501.3B, of which 23.4B are active per token
- Quantisation
- UD-Q4_K_XL
- Weights on disk
- 307 GB
- KV cache
- 1.3 GB at 32k context — 43 of the 52 layers use a 512-token window and stop growing there; the other 9 cache the whole context, so the full 1M window costs about 39 GB. The four-bit build is an unsloth dynamic quant, which allocates bits differently from a plain Q4_K_M; Reflection publishes BF16, FP8 and NVFP4 only. How weights and cache add up.
- Maximum context
- 1M tokens
- Licence
- Apache 2.0
- Sources
- reflection/Beam-1.0-501B-A23B, unsloth/Beam-1.0-501B-A23B-GGUF